AI Paper Atlas

Research map / 3D vision / Depth and stereo

Multi-view Gaze Target Estimation

Qiaomu Miao, Vivek Raju Golani, Jingyi Xu, Progga Paromita Dutta, Minh Hoai, Dimitris Samaras

ICCV 2025 · 3D vision › Depth and stereo · approach: Learned stereo with specialized sensors · 5 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add uncertainty score prediction (ablation)single-view GTE baseline · MVGT (reference: head visible, target visible)normalized L2 distance: 0.151 -> 0.145Improved
add Head Information Aggregation (HIA) (ablation)sigma model · MVGT (reference: head visible, target visible)normalized L2 distance: 0.145 -> 0.135Improved
add Uncertainty-based Gaze Selection (UGS) (ablation)sigma + HIA · MVGT (reference: head visible, target not visible)normalized L2 distance: 0.133 -> 0.123Improved
add Epipolar-based Scene Attention (ESA) (ablation)sigma + HIA + UGS · MVGT (reference: head not visible, target visible)normalized L2 distance: 0.170 -> 0.161Improved
add ESA (ablation)sigma + HIA + UGS · MVGT (reference: head not visible, target not visible)normalized L2 distance: 0.152 -> 0.152No clear effect
add HIA (in/out prediction, head not visible, target visible) (ablation)sigma model · MVGTAP: 0.756 -> 0.821Improved
feed relative camera rotation into HIA (ablation)HIA · MVGT (reference: head not visible, target visible)AP: 0.776 -> 0.821Improved
feed reference head image into HIA (ablation)HIA · MVGT (reference: head visible, target not visible)AP: 0.886 -> 0.897Improved
concatenate rather than multiply camera rotation in HIA (ablation)Ours · MVGT (reference: head not visible, target not visible)normalized L2 distance: 0.160 -> 0.152Improved
train the gaze backbone instead of freezing it (ablation)Ours · MVGT (reference: head not visible, target visible)normalized L2 distance: 0.193 -> 0.161Improved
apply UGS on samples with initial angular error above 30 degrees (ablation)Ours · MVGT large-error samplesnormalized L2 distance: 0.228 -> 0.200Improved
add ESA on occluded primary-view samples (ablation)Ours · MVGT occluded (reference: head visible, target visible)normalized L2 distance: 0.161 -> 0.150Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy